Future Prediction for Nurse Care Activities Using Deep Learning Based Multi-Label Classification
Md. Golam Rasul, Wasim Akram, Sayeda Fatema Tuj Zohura, Tanjila Alam Sathi, Lutfun Nahar Lota · 2024
Near future activity prediction from the previous activities, in health care system can be useful for health care management. Although a lot of dataset containing accelerometer data is available, accelerometer data with care record data of nurse care activities was explored by 4 Nurse Care Activity Recognition Challenge. The dataset is very challenging and complex because of time skewness, overlaps, data imbalance. Moreover, the nurse care challenge represented the data as a multi-label classification problem for identifying future activities. Traditional machine learning and deep learning models cannot be directly used for multi-label classification. In this paper, we configured a deep learning model for multi-label classification with necessary pre-processing. Average 63% accuracy was achieved outperforming the baseline scores given by the challenge whereas our model was evaluated with the test results of 87% accuracy and 11% F1-score [ 15 ].